A multi-disciplinary investigation of the AFEN Slide: the relationship between contourites and submarine landslides

Abstract Contourite drifts are sediment deposits formed by ocean bottom currents on continental slopes worldwide. Although it has become increasingly apparent that contourites are often prone to slope failure, the physical controls on slope instability remain unclear. This study presents high-resolution sedimentological, geochemical and geotechnical analyses of sediments to better understand the physical controls on slope failure that occurred within a sheeted contourite drift within the Faroe–Shetland Channel. We aim to identify and characterize the failure plane of the late Quaternary landslide (the AFEN Slide), and explain its location within the sheeted drift stratigraphy. The analyses reveal abrupt lithological contrasts characterized by distinct changes in physical, geochemical and geotechnical properties. Our findings indicate that the AFEN Slide likely initiated along a distinct lithological interface, between overlying sandy contouritic sediments and softer underlying mud-rich sediments. These lithological contrasts are interpreted to relate to climatically controlled variations in sediment input and bottom current intensity. Similar lithological contrasts are likely to be common within contourite drifts at many other oceanic gateways worldwide; hence our findings are likely to apply more widely. As we demonstrate here, recognition of such contrasts requires multi-disciplinary data over the depth range of stratigraphy that is potentially prone to slope failure.

Thermohaline-driven ocean bottom currents create sedimentary accumulations called contourites that are found along the world's continental margins (e.g. McCave and Tucholke 1986;Rebesco and Stow 2001;Stow et al. 2002). Contourites can cover extremely large areas (from ,100 km 2 to .100 000 km 2 ), forming a variety of depositional geometries that include elongated, mounded, sheeted, channelized and mixed drift systems (Faugères et al. 1999;Rebesco and Stow 2001;Stow et al. 2002;Faugères and Stow 2008). It has become increasingly apparent that contourite drifts are prone to slope instability (Laberg and Camerlenghi 2008), with submarine landslides recognized in a wide range of locations affected by bottom currents (Table 1).
The affinity of contourite drifts for slope failure can be linked in part to deposit morphology ( Fig. 1, Table 1). In some locations, contour-parallel currents modify the continental slope profile, creating mounded accumulations of sediment that are thicker and steeper than those on slopes unaffected by bottom currents (Laberg and Camerlenghi 2008;Rebesco et al. 2014). Factors such as sediment supply, intensity and location of currents, and sea-level and climatic changes control the presence or absence, location, growth and morphology of contourites (Faugères and Stow 2008;Rebesco et al. 2014). A number of compound morphological effects have been implicated as preconditioning and/or triggering mechanisms for slope instability. These include (1) slope over-steepening due to rapid sediment accumulation (A, Fig. 1) or due to erosion by vigorous along-slope currents (B, Fig. 1) and (2) loading resulting from differential sediment accumulation (C, Fig. 1). These effects occur particularly where contourites form as mounded accumulations (Laberg and Camerlenghi 2008;Prieto et al. 2016;Miramontes et al. 2018). However, submarine landslides, some of which include the largest on our planet (e.g. Storegga; Bryn et al. 2005a), often occur  Mounded drifts Under-cutting; weak layer Iwai et al. (2002), Volpi et al. (2003Volpi et al. ( , 2011 Slide volume, seabed gradient and sediment accumulation rate are given where available. Main controls of slope failure are listed where they are known or discussed in the literature. Submarine slope failure in contourites within contourite drifts on very low angle (,2°) slopes (e.g. Hühnerbach et al. 2004). Another explanation for slope instability in contourite drifts, therefore, relates to specific compositional and geotechnical properties of contourites ( Fig. 1, Table 1; Lindberg et al. 2004;Kvalstad et al. 2005). Plausible controls include prominent layers within the slope stratigraphy ( Fig. 1) that may feature a lower peak or post-peak shear strength than over-and underlying strata, such as (i) laterally extensive (sometimes cm-thin) homogeneous layers of weaker, sensitive material that is prone to sudden strength loss ( (Laberg and Camerlenghi 2008;Ai et al. 2014). Another plausible control relates to lithological and/or geotechnical contrasts within a depositional sequence that may result from rapid changes in current regime, sediment input or type (e.g. Rashid et al. 2017; see (iii) on Fig. 1). Detailed sedimentological and geotechnical studies of landslides within contourites are scarce (Baeten et al. 2013;Miramontes et al. 2018), and there is still much uncertainty as to which specific aspects act as the dominant control on slope instability. Many studies rely solely upon remote geophysical data for landslide characterization and, if sediment cores are acquired, they typically do not penetrate to the failure plane (which may be tens to hundreds of metres below the seafloor; Talling et al. 2014). Such cores also tend to focus on characterization of the failed landslide mass, rather than targeting sediments from adjacent undisturbed slopes. Targeting the undisturbed sediments of the adjacent slopes, including those stratigraphically equivalent to the failure plane of the landslide, however, is necessary in order to identify and characterize the material along which the landslide initiated, as these are usually removed or remoulded during failure. It is of critical importance to be able to identify sediments that are prone to failure in order to perform reliable slope stability assessments Vardy et al. 2012).

Aims
Here, we present a detailed characterization of a bedding-parallel, cohesive submarine landslide  Fig. 1. Key characteristics of contourites that favour the formation of submarine landslides. Morphological controls: (A) over-steepening, (B) erosion, (C) sediment loading; stratigraphic controls: (i) laterally extensive sensitive clay layers that are prone to sudden strength loss, possible shear strength depth profiles are shown as black; dark grey, dashed and light grey, dotted lines; (ii) thick accumulation of sandy layers that can accommodate excess pore pressure due to high sedimentation rates; (iii) distinct lithological and/or geotechnical interfaces. Contourite depositional system adopted from Hernández-Molina et al. (2008). (called the AFEN Slide), which occurred within a low angle (,2.5°) laterally extensive sheeted contourite drift, based on physical, geochemical, sedimentological and geotechnical analyses. We focus on a sediment core targeted to sample the prelandslide sedimentary sequence, including sediments that correlate stratigraphically with the failure plane located further upslope. Based on centimetreresolution characterization of these deposits we address the following questions: first, what is the nature of the undisturbed sediment and do material heterogeneities explain the location of the failure plane? As many aspects of cohesive landslides appear to be scale invariant, this study of a relatively small landslide may provide key insights into our understanding of much larger ones (Micallef et al. 2008;Chaytor et al. 2009;Baeten et al. 2013;Casas et al. 2016;Kuhlmann et al. 2017;Clare et al. 2018). Second, what causes the observed heterogeneities within the stratigraphy? We explore how climatic changes and ocean circulation may play a key role in governing not just the failure plane depth, but also influence the timing of slope failure. Finally, we discuss the implications of climatically controlled sediment supply and deep ocean circulation for pre-conditioning slope instability in contourite depositional systems in oceanic gateways, which are narrow, deep passages connecting adjacent basins, elsewhere in the world.

Regional setting
Geological and morphological setting. The study area lies on the eastern margin of the Faroe-Shetland Channel, which is located north of Scotland, extending over 400 km between the Wyville-Thomson Ridge and the Norwegian Basin (Fig. 2). The Faroe-Shetland Channel is a narrow basin, measuring 250 km at its widest in the NE and less than 130 km in the SW. The channel closely follows the trend of the regional NE-SW structural lineaments, and one of the NW-SE transfer zones (Victory Transfer Zone) passes close to the study area (Rumph et al. 1993;Wilson et al. 2004). The Faroe-Shetland Channel is the present-day expression of the Faroe-Shetland Basin that can be dated back to the Late Paleozoic (e.g. Rumph et al. 1993). Basin formation was probably initiated during the Devonian, while the main rift phase occurred during Cretaceous times (Dean et al. 1999;Roberts et al. 1999). Although localized extension continued until the early to mid-Paleocene (Smallwood and Gill 2002), more or less continuous post-rift subsidence predominated throughout the Cenozoic (Turner and Scrutton 1993). This subsidence was interrupted at various stages by contractional deformation (Ritchie et al. 2003(Ritchie et al. , 2008Johnson et al. 2005;Stoker et al. 2005) and regional uplift and tilting (Andersen et al. 2000;Smallwood and Gill 2002;Stoker et al. 2002Stoker et al. , 2005. Following Late Paleocene uplift, the Faroe-Shetland Channel subsided about 2000 m, with present-day water depths of 1700 m in the NE and 1000 m in the SW, and slope angles between 1°and 3°flanking the eastern channel margin (Stoker et al. 1998;Andersen et al. 2000;Smallwood and Gill 2002). The channel forms an important oceanic gateway, exchanging water masses between the North Atlantic and the Norwegian Sea (Broecker and Denton 1990;Rahmstorf 2002) since at least the Early Oligocene (Davies et al. 2001).
Oceanography and palaeoceanography. In general, the present-day oceanography in the Faroe-Shetland Channel consists of warm surface water moving towards the NE, and cold bottom water, generating relatively strong, erosive bottom currents (with velocities in the range between ,0.3 and .1.0 m s −1 ; Masson et al. 2004), moving towards the SW (Fig. 2a;Saunders 1990;Turrell et al. 1999;Rasmussen et al. 2002). Five distinct water masses can be recognized based on their salinity and temperature characteristics (Turrell et al. 1999). Two distinct surface water masses transport warm water from the North Atlantic into the channel. North Atlantic Water (NAW) flows northward from the Rockall Trough (Turrell et al. 1999), while Modified North Atlantic Water (MNAW) flows clockwise around the Faroe Islands before turning northward in the Faroe-Shetland Channel (Saunders 1990). These surface waters typically occupy the upper 200-400 m of the water column (Turrell et al. 1999). Arctic Intermediate Water (AIM) flows anticlockwise along the southern edge of the Norwegian Basin and around the Faroe-Shetland Channel, typically between 400 and 600 m water depth (Blindheim 1990). At the base of the channel (usually below 600 m water depth), the Norwegian Sea Arctic Intermediate Water (NSAIW) and the Faroe-Shetland Channel Bottom Water (FSCBW) are funnelled along the Faroe-Shetland Channel towards the south (Turrell et al. 1999) and flow along the Faroe Bank Channel into the Atlantic (Saunders 1990). A small portion of the cold bottom water flows across the western end of the Wyville-Thomson Ridge south into the Rockall Trough (Stow and Holbrook 1984). The velocity of these water masses is variable, both across the channel and over time. The average along-slope velocities, mainly directed NE, of around 0.2-0.25 m s −1 were measured at around 500-700 m water depth (Van Raaphorst et al. 2001;Bonnin et al. 2002) and velocities over .1.0 m s −1 associated with SW-directed bottom currents were inferred from observed bedforms . Periodic changes in salinity and temperature cause shifts of the boundaries between water masses on timescales from decades to hours (Turrell et al. 1999). Since the Last Glacial Maximum (LGM), when bottom and surface currents were weak, eight distinct changes in the surface and bottom current regime were identified, which are related to the changes in climatic conditions (Rasmussen et al. 2002). Climatic and palaeoceanographic changes also reportedly caused strong cyclical variation in sediment accumulation (with up to 30 cm ka −1 along the Faroe Drift and up to 10 cm ka −1 along the West Shetland Drift; Rasmussen et al. 1996Rasmussen et al. , 1998Knutz and Cartwright 2004;Nielsen et al. 2007).

Contourite deposits in the Faroe-Shetland Channel
The regional oceanography has controlled the depositional architecture of the slope sediments, creating elongated mounded contourite drifts at the base of the slope (to the NE of the AFEN Slide) and sheeted contourite drifts in the slide area Hohbein and Cartwright 2006). These sheeted drifts are characterized by parallel, laterally continuous reflectors on seismic profiles (Masson 2001). These reflectors can be traced over more than 50 km below the seafloor of the Faroe-Shetland Channel, which emphasizes the regional scale of bottom current activity and sheeted contourite drift accumulation (Stoker et al. 1998).

The AFEN Slide
The AFEN Slide was first identified in 1996, during an environmental survey for the Atlantic Frontiers Environmental Network in the region (Wilson et al. 2004). The slide is interpreted as a four-stage retrogressive landslide that occurred NW of the Shetland Islands (UK) at water depths of 830-1120 m on a slope varying from approximately 0.7°to about 2.5° (Wilson et al. 2003(Wilson et al. , 2004Fig. 2b). The total length from the head scarp to the toe of the lobe is over 12 km, and the maximum width is around 4.5 km. The slide involved c. 200 × 10 6 m 3 of sediment and the slide debris has a maximum thickness of 20 m, averaging between 5 and 10 m (Wilson et al. 2004). Radiocarbon dating and biostratigraphy from the slide suggest that the first stage took place around 16-13 ka BP and the later retrogressive phases after 5.8 ka BP and prior to 2.8 ka BP (Wilson et al. 2004). Initial studies, based on highresolution seismic data and sediment cores that did not penetrate the base of the slide, inferred that the failure plane comprised well-sorted contourite sands, which may liquefy during an earthquake (e.g. 10 000-year return period earthquake; Jackson et al. 2004). This hypothesis was supported by the presence of a buried slide, which appears to have occurred under similar physiographic conditions (Masson 2001;Wilson et al. 2003Wilson et al. , 2004. Such wellsorted contourite sands were not found by Madhusudhan et al. (2017), who analysed a new sediment core (64PE391-01) that penetrated through the full extent of the deposits from the second stage of the landslide (Fig. 2c). Instead, they proposed progressive failure of geotechnically sensitive clays or liquefaction of silt layers. None of these previous cores sampled undisturbed material that corresponds stratigraphically with the failure plane.

Data and methods
Core 64PE391-04, which is the focus of this present study, was obtained during the RV Pelagia cruise 64PE391 in 2014 using a piston corer. The core was sampled within the AFEN Slide area, at a water depth of 945 m. It was targeted to sample undisturbed sediments, i.e. those characterized on seismic data by continuous reflectors and avoiding acoustically transparent, chaotic or disrupted seismic units and areas of hummocky seafloor texture likely indicative of slope failure (Shipp et al. 2011;Fig. 2). Figure 2 shows the location of core 64PE391-04 on the deep tow boomer seismic profile, which has a maximum theoretical vertical resolution of 0.5 m, with a penetration of 100 ms, and was obtained from the BGS 00/02 survey (Wilson et al. 2005). The core recovered 11.49 m of sediment in a 15 m core barrel and was kept in refrigerated storage at the British Ocean Sediment Core Facility (BOSCORF), UK, prior to study.

Physical properties analysis
A Geotek MSCL-S (Standard) multi-sensor core logger, based at BOSCORF, was used to measure P-wave velocity, gamma-ray bulk density, electrical resistivity, magnetic susceptibility and fractional porosity, which is derived from the measured sediment density at 1 cm intervals on split cores (Fig. 3). MSCL is a commonly used, non-destructive tool that allows the recognition of subtle changes in sediment physical properties. The data are commonly used for correlation between cores, and calibration of seismic data using P-wave velocity. Density serves as an effective proxy for changes in sediment lithology and is used for the calculation of fractional porosity (Gunn and Best 1998). Core images were obtained using the BOSCORF Geotek MSCL-CIS (Core Imaging System), which enables the acquisition of precise depth-registered images that can be correlated with the other datasets.

Geochemical analysis
Micro-XRF (X-ray fluorescence) core scanning was used to determine the geochemical composition of the sediment (ITRAX™ COX Ltd at BOSCORF; Croudace et al. 2006) at a spatial resolution of 1 cm. ITRAX scanning is a useful, rapid, non-destructive, high-resolution scanning technique that is widely used in earth and environmental sciences (Croudace and Rothwell 2015). This method enables the measurement of element intensities, such as Ca and Sr, which correlate well with the carbonate content, or Fe, Ti and K, which are related to the siliciclastic components, and vary directly with the terrigenous sediment input (e.g. Röhl and Abrams 2000;Hepp et al. 2006). ITRAX data represent a semiquantitative analysis of the relative element abundances down-core. Data are expressed as counts per second (cps) and are presented as log ratios, which are accepted as a more accurate estimation of element concentrations. In addition, all XRF data are shown as log ratios of two elements, in order to show element concentrations more accurately and minimize matrix effects inherent to XRF (Weltje and Tjallingii 2008). Ca/Sr, Ca/Fe and Fe/K have been selected, as these element ratios have been shown to reflect changes in sea-level and temperature, sediment supply, and have been applied in climate studies (see Croudace and Rothwell 2015). In addition to geochemical composition, the ITRAX instrument provided X-radiographs. X-radiographs are digital images of the internal structure and physical property changes within a split core section that are obtained using optical and radiographic line cameras.

Grain-size distribution
Grain-size analysis was carried out at 10 cm depth intervals for sediments of Units 2, 3 and 4 (see results for definition), following the procedures in Rothwell and Rack (2006). The sediment was sieved to remove particles larger than 2 mm before the sample was dispersed in a 1 l mixing chamber by shaking it for 24 h. The dispersed sediment was circulated through a Malvern Mastersizer 3000 for 120 s over which time 12 measurements are taken and then averaged to obtain the grain-size distribution.

Geotechnical analyses
Water content and fall cone measurements were carried out at 10 cm intervals (BSI 1990;BSI 2004). Measurements of water content could be used as a first-order approximation of the sediment's shear strength and compressibility (i.e. higher water content is related to low shear strength and high compressibility). An 80 g 30°fall cone was used on the split cores, regardless of the grain size and whether the tested material was considered to be saturated or not. The undrained shear strength was calculated from the fall cone measurements assuming all tests were carried out on saturated clays. Subsamples were taken for subsequent direct shear and oedometric tests.
Static, drained shear test. Direct shear experiments were carried out to compare the drained shear strength of prominent layers, identified from downcore logging, grain-size distribution and standard geotechnical data. Cylindrical, undisturbed samples (c. 5 cm 2 , 2.5 cm height) of intact samples were placed in the shear apparatus and consolidated via a vertical ram to in situ normal stress (σ n ). The sample was consolidated until the sample height was constant (or min. 24 h), so that the sample is assumed to be fully drained and the applied σ n is approximately equal to the effective normal stress (σ′ n ). The effective normal stress is the difference between the normal stress and the pore water pressure, u (σ′ n = σ nu; Terzaghi 1925). Shearing occurs on a predefined plane, perpendicular to the vertical ram that exerts the normal stress. The shear displacement for each experiment was 9.5 mm at a shear rate of 0.008 mm min −1 . This shear rate was slow enough to allow constant drainage during shearing (Deutsches  Fig. 3. Summary of sediment core analyses (64PE391-04), including visual sedimentary, physical properties (multi-sensor core logging) and geochemical (ITRAX X-ray fluorescence) core-log data, and geotechnical data (water content, drained and undrained shear strength). Units 1-5 are outlined.
Institut für Normung 2002). Samples were taken from around 7 m core depth, which corresponds to around 18 m below seafloor (assuming around 10 m of sediment was removed during the failure). The samples were sheared at a normal stress of 170 kPa, simulating the effective hydrostatic vertical overburden stress (σ′ v0 ) acting at around 18 m below seafloor (m bsf ) assuming an average sediment effective unit weight (γ′) of 9.5 kN m −3 .
Oedometer test. One-dimension consolidation tests were performed on selected undisturbed core samples (c. 20 cm 2 , 1.9 cm height) in order to measure and compare their permeability and consolidation parameters. The measured initial porosity (n), coefficient of compression (c v ) and permeability (k) can be used to make assumptions regarding the sediments' potential to build excess pore pressure. Incremental loading and unloading of 1 kPa to 7100 kPa stress were applied onto the sediment and the resulting displacement (change in volume) was measured. Each load was applied gradually and left until the displacement stabilized or primary consolidation was completed. Consolidation and permeability parameters were calculated from the settlement characteristics of the sediment using standard equations (Powrie 2013).

Data analysis
Physical and geochemical properties were compared using non-parametric tests that compare two unpaired groups of data and compute p-values testing the null hypothesis of two groups having the same distribution. The data were analysed for the discrepancy between the mean ranks of two groups (Mann-Whitney test) and for their varying cumulative distribution (Kolmogorov-Smirnov test) (Sheskin 2011). The significance level for both tests was set to 0.05 (Fisher 1926).

Results
Piston core 64PE391-04 was obtained about 750 m down-slope from where the sediment ramped up the failure plane on to the seabed (failure Stage 1, Wilson et al. 2004 ; Fig. 2c). The deep-tow boomer reflection seismic data indicate that the core penetrated the pre-landslide sediments, including those stratigraphically equivalent to the failure plane of the slide. Based on the newly obtained data, we identify five main lithological units within the sediment core, which we now characterize using results from visual sediment core logging, particle size distribution, X-ray scanning, and continuous physical properties (MSCL) and geochemical (micro-XRF) measurements (see summary in Figs 3 & 4). In addition, we present a geotechnical characterization of the recovered sediment based on water content and fall cone analyses, as well as direct shear (DS) and oedometer tests.
Visual sedimentary logging and grain-size analyses indicate that the general lithology is bioturbated silty clay to clayey silt with a number of sandy silt and silty sand layers, consistent with previous analyses of sediment cores from the area (Madhusudhan et al. 2017). Sandy layers are found only in the upper part of the core (above 7.3 m depth). The lithology in the lower part of the core is generally homogeneous with an absence of sand.

Multi-Sensor Core Logger (MSCL) data
Down-core logging data show an abrupt and distinct change in physical properties at around 7.3 m depth, as well as more subtle variations that enabled demarcation of the five sediment units ( Fig. 3; Table 2). Unit 1 is largely indiscernible from Unit 2 based on physical properties, but does have much lower magnetic susceptibility. The sediments above the abrupt contact at 7.3 m (Units 2 and 3) are generally characterized by high relative P-wave velocities, gamma-ray densities, electrical resistivity, and low relative values of fractional porosity (on average under 0.5). Unit 3 shows the highest electrical resistivity and gamma-ray densities in the core; hence, it is demarcated as an individual unit, rather than being subsumed within Unit 2. In the sediments immediately below 7.3 m (Unit 4), the most marked step in physical properties is observed, including a reduction in gamma-ray density from 2.0 to 1.7 g cm −3 , and an increase in fractional porosity from approximately 0.45 to .0.55. Such a marked change was not observed in the magnetic susceptibility this side of the contact either; however, the signal is generally more erratic above and less variable below (Fig. 3). Below the contact at 7.3 m, P-wave velocity, gamma-ray density and electrical resistivity gradually increase down-core (inversely mirroring a steady decrease in fractional porosity) until the start of Unit 5, which is marked by a sharp increase in magnetic susceptibility (from ,70 to .165 m 3 kg −1 ), and a subtle increase in average P-wave velocity and gamma-ray density (Fig. 3).

Micro-X-ray fluorescence (XRF) data
Distinct changes in geochemistry are also observed from the micro-XRF analysis between the sediment units (Figs 3 & 5), which correspond to very similar depths (+0.3 m) where physical property changes are noted. The first-order observations are of: (i) a step in Fe/K, Ca/Fe and Ca/Sr elemental ratios between 7.1 and 7.3 m (i.e. straddling Unit 2/3/4 contacts); and (ii) a switch from more variable (noisy) elemental ratios above 7.1-7.3 m (Units 2 and 3), with centimetre-scale variations in geochemical composition, to less noisy ratios below (Unit 4). Below Unit 4, variations in elemental ratios are also observed, supporting the demarcation of Unit 5. Cross-plotting of the elemental ratios (Fig. 6) supports the demarcation of the five identified sediment units, as well as illustrating the range in variability between each unit (e.g. a large spread of values in Unit 2, compared to Unit 4). Figure 7 summarizes grain-size distribution data for core section 64PE391-04-D (6.5-7.7 m depth), which include sediments from Units 2, 3 and 4. The data illustrate the change in composition at around 7.3 m depth. Unit 4 (below 7.3 m depth) is characterized by a higher silt content in comparison to overlying sediments. Unit 3 is recognized as a sandy silt layer, and the sampled sediments of Unit 2 show a switch from sandy silt to clayey silt, which supports the distinct changes in lithology seen in the visual core log.

Geotechnical data
A distinct change in water content can be observed, which increases from around 30% to over 60% at 7.3 m depth (i.e. at the contact between Units 3 and 4; Fig. 3). Unit 1 has a slightly higher water content than Unit 2 (more or less constant 30%). Units 4 and 5 are characterized by decreasing water content. A distinct change in the undrained shear strength is not observed, although the scatter is greater in the upper part of the core (Units 2 and 3). Individual outliers (.100 kPa) are related to dropstones or mud clasts.
A summary of the key sample parameters and test results of the direct shear and oedometer tests are given in Table 3. The peak drained shear strength of Units 3 and 4 are shown in Figure 3 (indicated by red crosses). It can be seen that Unit 3 encompasses a higher peak shear strength (173 kPa) than Unit 4 (109 kPa). Typical porosity (n) v. applied normal stress (σ n ) is shown in Figure 8. It is apparent that porosity decreases with increasing normal stress and increases slightly during the rebound phase. Unit 3 has a lower initial porosity, and higher permeability (k) and compressibility (c v ) than Unit 4.

Discussion
The recovered slope sediment obtained from core 64PE391-04 is characterized by a distinct step change in both physical and geochemical properties between around 7.1 and 7.3 m depth, as well as a distinct high-density contrast at that depth that was recorded by X-ray imaging (Figs 3 & 4). These transitions are related to an abrupt change in lithology from a thick relatively homogeneous clayey silt, silty clay unit (Unit 4; Figs 3 & 5) to an overlying 25 cm-thick sandy silt layer (Unit 3; Figs 3 & 5). The depth of this distinct change matches well with the seismostratigraphic horizon that is equivalent to the main failure plane outlined in the deep-tow boomer reflection seismic data (assuming a seismic velocity of 1600 m s −1 ; Wilson et al. 2004), which is supported by the available MSCL data.
The sediment above this distinct interface is characterized by slightly higher P-wave velocities and gamma-ray densities, as well as a lower fractional porosity than would be expected for continental slope sediments ( Fig. 3; Hamilton 1970). Small cracks were recorded by X-ray imaging, but are limited to parts of Unit 2 (Fig. 4). These observations could be related to a slight compaction of the sediment, for example, due to compression by the partially confined landslide debris above the sediment ramp ( Fig. 2c; Fig. 4. Inferred location of the main failure plane based on down-core logging and deep-tow boomer reflection seismic data. Units 1-5 are outlined. Vertical error in the failure plane delineation, resulting from the vertical resolution of the seismic data is indicated by grey lines (+50 cm from the inferred failure plane). Core images and X-radiographs from the inferred failure plane and cracks in Unit 2 are also shown.

Steady interstadial deposition
Submarine slope failure in contourites 10 m missing sediment sequence at the 64PE391-04 core location (Fig. 2), whose removal could have disturbed the slope sediments. The potential deformation, however, is not resolved in the seismic data, and the distinct change at around 7.1-7.3 m depth is not limited to the physical properties, but is also noted in the geochemical properties. We therefore infer that although the sediment might have been slightly deformed, it probably did not move (no sliding motion) and the stratigraphy was not altered.
Lithological contrasts appear to play a key role in dictating the location of the failure plane Wilson et al. (2004) previously suggested that the AFEN Slide could have initiated along a sandy contouritic layer embedded within the slope stratigraphy, but were unable to sample deep enough to show its occurrence. Our deeper core now shows that this hypothesis may be plausible, given the presence of Unit 3. Although this unit was not identified as a contourite in the seismic data (Fig. 2c, Wilson et al. 2004), we interpret it as a sheeted sandy contourite drift. This assumption is considered reasonable as the vertical resolution of the seismic data (0.5 m; Wilson et al. 2005) might be too low to register this 25 cm-thick layer. Furthermore, we also show that there is much greater lithological heterogeneity (based on physical properties and geochemistry) within these sheeted drifts than has been previously documented, aside from simply variations in grain size. Without detailed geochemical and physical properties data, this abrupt lithological change would not have been identified. Abrupt lithological changes (such as between Units 3 and 4) may instead play a key role in defining the location of the failure plane. Unfortunately, the vertical resolution of the existing seismic data does not enable us to categorically determine whether the failure plane should correspond to the contact of Units 3/4 or 2/3. Although varying the assumed seismic velocity within reasonable ranges for sediments only results in a vertical offset of 0.5 m, the failure plane falls within the depth window that includes the interfaces between Units 2/3 and 3/4 (Fig. 4). Wilson et al. (2004) implicated sandy contouritic sediments as potential 'weak layers' (i.e. Units 2/3 scenario) because of their potential to generate excess pore pressures when bound by an overlying lower permeability unit. This is a reasonable suggestion; however, the fractional porosity data indicate that the sand-rich Unit 3 instead features slightly lower porosity than the overlying sediments, while the underlying mud-rich sediments (Unit 4) have an even higher porosity. This observation is supported by water content data, which show the highest values in the mud-rich Unit 4 and abruptly   5.2 × 10 −4 7.6 × 10 −5 k (m s −1 ) 4.3 × 10 −7 7.8 × 10 −8 LL, liquid limit; PL, plastic limit; γ′, effective unit weight; σ′ n , effective normal stress; τ peak , peak shear strength; n, porosity; c v , compressibility; k, permeability.
decrease at the interface to Unit 3. Oedometer tests carried out on undisturbed samples from Units 3 and 4 reveal a higher initial porosity and lower compressibility of Unit 4. This relationship is in contrast to an established empirical relationship between coarser grain size and greater porosity (or larger pore size; Ren and Santamarina 2018). This apparent contradiction is explained by the presence of detrital clay that fills in pore spaces between sand grains (Unit 3), whereas the relatively open structure of the underlying muddier deposits (Unit 4) explains their higher relative porosity (Marion et al. 1992;Revil and Cathles III 1999). In contrast to porosity, however, permeability is found to be higher in the sand-rich sediments (Unit 3; Table 3). Considering the higher permeability and compressibility of Unit 3, it is possible for excess pore pressure to accumulate within the sandy contouritic sediments (e.g. during an earthquake). Although this observation would support the 'weak layer' hypothesis, it has to be noted that the water content is actually higher in Unit 4 and abruptly drops at the interface with Unit 3, instead of increasing within the layer. Another noticeable observation is the difference in shear strength between Units 3 and 4. Both drained and undrained shear strength are lower in the mud-rich Unit 4, which can be related to the higher water content and the lack of sandy material within the unit. Taking all these observations into account, we suggest that it is possible that a failure plane could generate at an interface where sand overlies finer-grained cohesive sediments. The high water content and lower shear strength of the finegrained material could allow the overlying sediment to slide on top of it. We are unable to be more absolute on the failure depth, but we have demonstrated that variability in sheeted drifts can also include abrupt whole-scale changes in sediment properties, as well as the presence of thin coarser units, which have traditionally been invoked to explain beddingparallel failures in contourite sheeted drifts (Laberg and Camerlenghi 2008). Such variability may not necessarily be expected based on the available seismic data alone.
Climate change is a likely control on creating failure-prone lithological contrasts Down-core changes in Ca/Sr ratios have been successfully related to variations in sea-level and water temperature (through integration with oxygen isotope curves and biostratigraphy), wherein high Ca/Sr ratios are indicative of ice-rafted debris and changes from colder to warmer conditions (e.g. Smith et al. 1979;Thomson et al. 2004;Hodell et al. 2008). High Fe/K ratios and low Ca/Fe, on the other hand, have been related to colder periods (Kuijpers et al. 2003;Perez et al. 2016). The increased Ca/Sr ratio above 7.6 m depth could therefore indicate a stronger meltwater flux, carrying icerafted debris into the channel, while the changes in Fe/K and Ca/Fe ratios at 7.1-7.3 m are also interpreted to indicate a switch from cold conditions (Unit 4) to warmer conditions (Units 2/3). This switch was coincident with a transition from finergrained, stable sedimentation to a more variable regime with pulsed influxes of coarser material. Given the existing knowledge about the timing of the AFEN Slide (Unit 1 should post-date 2.8-5.8 ka BP, while the pre-failure sediments must be older than 16 ka BP; Wilson et al. 2004), this transition fits within a time window that includes the switch from the LGM (18 ka BP) to post-glacial conditions. Glacial conditions would have seen sediment largely locked up in ice sheets, while the melt-out during the immediate postglacial window involved pulses of fineand coarser-grained sediment. The nearby Faroe-Shetland Channel is the main oceanic gateway between the North Atlantic and the Norwegian Sea (Broecker and Denton 1990;Rahmstorf 2002), where a direct relation exists between ocean circulation and climate. Rapid changes in the exchange of water masses between the NE Atlantic and the Norwegian Sea occurred following the LGM at 18 ka BP (Rasmussen et al. 2002), which would have compounded the abruptness of a switch in sediment transfer. We therefore suggest that the abrupt change in physical properties and geochemistry may relate to this climatic transition.
Previous studies have investigated the role of climate change on submarine landslides, primarily focusing on their timing. A number of early studies suggested that submarine landslides, particularly in higher latitudes, may be more likely during sea-level lowstands. Recent work, however, has suggested that there is no clear statistical relationship or at least that there are too few observations to be confident (e.g. Maslin et al. 2004;Brothers et al. 2013;Urlaub et al. 2013Urlaub et al. , 2014Pope et al. 2015). Indeed, recent work has shown that such margins may feature many more late Holocene submarine landslides than previously thought (Normandeau et al. 2019). Proving a clear link between submarine landslides and sea-level or climate change is most likely complicated by a range of factors, including time lags in offshore sediment transport, residence times of excess pore pressures following periods of rapid sediment accumulation, local sea-level changes (e.g. isostatic rebound following glaciations) and other factors (Masson et al. 2006;Urgeles and Camerlenghi 2013;Talling et al. 2014). Whether climate change has played any role in the timing of the slope failures at AFEN remains unclear; however, it may have played a key role in one aspect: the location of the failure plane. Our data indicate that the slope failure most likely initiated along a distinct lithological interface that is interpreted to relate to a switch in depositional regime: from cold and uniform to warm and variable depositional conditions. The close connection between thermohaline circulation, sea-level and temperature, and sediment supply in this region may explain why the switch in deposition was so rapid.

Broader implications for slope instability in contourites at climatically influenced ocean gateways
The origin of distinct lithological interfaces may arise in a variety of ways, and may be very common in contouritic sediments near ocean gateways where climatic changes may affect bottom current intensity (and thus control the grain-size that is transported; Faugères and Mulder 2011), as well as the type of sediment that is distributed by bottom currents (e.g. terrestrial and biogenic fluxes may vary during different climatic windows; Faugères et al. 1993;Maldonado et al. 2005). Such effects can be felt at a variety of latitudes, ranging from tropical to polar settings (e.g. Kuijpers et al. 2001;Principaud et al. 2015;Elger et al. 2017). Climate may play a key role in dictating the location of potential failure planes. While many previous studies have invoked dominantly geometric controls on slope failure in contourite drifts, our study contributes to a growing literature base that indicates that lithological interfaces may explain the strong affinity of contourite deposits to slope instability. We posit that in lowangle, sheeted contourite drifts, such as AFEN, it is such material interfaces that are most important for preconditioning slopes to failure.

Conclusions
The integration of physical properties and geochemical core-log data, grain-size distribution, and geotechnical data indicate that the AFEN Slide initiated along a distinct lithological interface within the slope stratigraphy, which matches the depth of the failure plane obtained from seismic data. This lithological interface correlates with the base of a 25 cm sandy contourite layer, overlying a thick, relatively homogeneous silty clay unit. Based on this highresolution multi-proxy analysis, it was possible to resolve small-scale material changes within the slope stratigraphy, which cannot be distinguished from seismic data alone (owing to the limited vertical resolution of 0.5 m). Integrating the core analyses with our knowledge about the current regime prevailing in the Faroe-Shetland Channel for the last 18 ka, it seems that climate change might precondition the location of failure initiation. This highlights the fact that in order to understand submarine landslide hazard, it is necessary to include information from all different scales, ranging from the smallscale high-resolution analysis of core material to the understanding of the regional oceanographic setting.
Acknowledgements The authors thank the British Geological Survey for the supply of the deep-tow boomer reflection seismic data and previous contributions (in particular David Tappin and David Long). Thanks are also given to the crew of the RV Pelagia for their efforts during data collection. We thank BOSCORF and its staff (S. MacLachan, M. Edwards and M. Charidemou) for their services in maintaining the cores and assisting with some of the analytical techniques, Achim Kopf for letting us use his geotechnical laboratory at MARUM to carry out the shear tests, and the National Infrastructure Laboratory, Southampton, for letting us conduct our oedometer tests. We acknowledge the constructive reviews by A. Cattaneo and U. Nicholson, and editor J. Mountjoy. Credit for the bathymetric metadata is given to ESRI, Garmin, GEBCO, NOAA NGDC, and other contributors. Author contributions RG: conceptualization (equal), data curation (supporting), investigation (equal), methodology (equal), validation (equal), visualization (lead), writingoriginal draft (lead), writingreview and editing (lead); MAC: conceptualization (equal), project administration (equal), supervision (lead), visualization (supporting), writingoriginal draft (equal), writingreview and editing (equal); JEH: data curation (equal), investigation (equal), methodology (equal), project administration (equal), supervision (supporting), validation (equal), writingreview and editing (supporting); MW: data curation (equal), investigation (equal), validation (equal), writingreview and editing (supporting); BNM: conceptualization (equal), investigation (equal), methodology (equal), supervision (equal), validation (equal), writingreview and editing (supporting); PJT: conceptualization (equal), funding acquisition (lead), project administration (equal), writingreview and editing (equal); KH: conceptualization (equal), funding acquisition (lead), supervision (equal), writingreview and editing (supporting).